Low-leakage DRAM structures using selective silicon epitaxial growth (SEG) on an insulating layer

Abstract

Low current leakage DRAM structures are achieved using a selective silicon epitaxial growth over an insulating layer on memory cell (device) areas. An insulating layer, that also serves as a stress-release layer, and a Si₃N₄ hard mask are patterned to leave portions over the memory cell areas. Shallow trenches are etched in the substrate and filled with a CVD oxide which is polished back to the hard mask to form shallow trench isolation (STI) around the memory cell areas. The hard mask is selectively removed to form recesses in the STI aligned over the memory cell areas exposing the underlying insulating layer. openings are etched in the insulating layer to provide a silicon-seed surface from which is grown a selective epitaxial layer extending over the insulating layer within the recesses. After growing a gate oxide on the epitaxial layer, FETs and DRAM capacitors can be formed on the epitaxial layer. The insulating layer under the epitaxial layer drastically reduces the capacitor leakage current and improves DRAM device performance. This self-aligning method also increases memory cell density, and is integratable into current DRAM processes to reduce cost.

Description

This is a division of patent application Ser. No. 09/697,946, filling date Oct. 30, 2000, A Method For Making Low-Leakage Dram Structures Using Selective Silicon Epitaxial Growth Growth (Seg) On An Insulating Layer, assigned to the same assignee as the present invention.

BACKGROUND OF THE INVENTION

(1) Field of the Invention

The present invention relates to the fabrication of integrated circuit devices on semiconductor substrates, and more particularly relates to a method for fabricating Dynamic Random Access Memory (DRAM) cells using selective silicon epitaxial growth over an insulating layer on the cell (device) areas. The method is particularly useful for reducing capacitor leakage currents and soft error due to Alpha particles on DRAM cells.

(2) Description of the Prior Art

Advances in the semiconductor process technologies have dramatically decreased the semiconductor device feature sizes and increased the circuit density on the integrated circuits on chips. One device type that has experienced a rapid increase in density is the array of memory cells on DRAM devices. Each memory cell consists of a single pass transistor (FET) and a storage capacitor. As the cell area decreases and the capacitance of the storage capacitor decreases, it becomes increasingly difficult to maintain sufficient charge on the capacitor due to the capacitor leakage current, and the refresh cycle time needed to maintain the charge on the capacitor becomes unacceptably short. Another problem is the natural presence of Alpha particles, which can generate electron-hole pairs resulting in soft errors in the more conventional DRAM capacitors in which their node contacts are made directly to the diffused junctions in the silicon substrate.

One method of reducing the leakage current and reducing soft error is to use a silicon-on-insulator (SOI). However, SOI technology is still too expensive and complicated for manufacturing. However, as devices are further diminished in size, the junction depths and well depths decrease proportionally. The use of a thin silicon epitaxial layer is required for future device generations to achieve these shallow device structures.

Several methods for making and using SOI have been described in the literature. For example, in U.S. Pat. No. 5,691,776 to Hebert et al. a method is described for forming field oxide regions by etching trenches in which a conformal silicon nitride (Si₃N₄) is deposited over the trenches. An opening is etched in the Si₃N₄ layer and a selective epitaxial growth (SEG) is used to partially fill the trenches. The SEG is then thermally oxidized to form the field oxide. In U.S. Pat. No. 5,686,343 to Lee, Lee isolates a semiconductor layer on an insulator by first forming an insulating layer on a silicon substrate, etching a window to the substrate, depositing an amorphous silicon layer that is annealed to form an epitaxial layer over the window. The epitaxial layer is patterned and a Si₃N₄ layer is deposited over the patterned epitaxial layer, and a thermal oxidation is used to oxidize the silicon in the window under the semiconductor layer. In U.S. Pat. No. 6,037,199 to Huang et al. an insulating layer is formed on a silicon substrate, an opening is formed in the insulator, and an amorphous silicon layer is deposited and annealed to form an epitaxial layer extending from the opening laterally over the insulating layer. The epitaxial layer is patterned over the insulating layer to form isolated silicon regions (islands) in which FETs are formed. In U.S. Pat. No. 5,763,314 to Chittipeddi a method is described for forming two separate selective epitaxial layers, having different dopant concentrations, on the same silicon substrate. The epitaxial layers are separated by a trench filled with an insulating material.

However, there is still a strong need in the semiconductor industry to provide DRAM cells with low capacitor-leakage currents and reduced Alpha soft errors while providing a process that is integratable into the current manufacturing process without significantly increasing manufacturing process complexity.

SUMMARY OF THE INVENTION

Therefore a principal object of this invention is to make DRAM cells with increased cell density while reducing capacitor leakage currents.

Another object of this invention is to reduce the leakage currents and soft error by using a silicon epitaxial layer over an insulating layer on which are formed the DRAM FETs and storage capacitors.

It is another object to integrate this novel DRAM cell into the current DRAM process to minimize manufacturing cost by integrating the selective silicon epitaxy on insulator without significantly increasing the processing steps.

Another objective of this invention by a first embodiment is to make a flat capacitor structure using this selective epitaxy DRAM process having low leakage currents.

Still another objective of this invention by a second embodiment is to make a stacked capacitor structure using this selective epitaxy DRAM process having low leakage currents.

In accordance with the objects of the present invention a method for fabricating dynamic random access memory (DRAM) cells on and in an epitaxial silicon layer formed over a first insulating layer on a semiconductor substrate is described. The method by a first embodiment begins by providing a P doped single-crystal silicon semiconductor substrate for N channel FETs. Alternatively an N doped substrate can be used if P channel FETs are desired. A first insulating layer that also serves as a stress-release layer is formed on the substrate. A hard-mask layer composed of Si₃N₄ is deposited on the first insulating layer. The hard mask is patterned to leave portions over the desired device areas. The hard mask and plasma etching are then used to etch shallow trenches in the substrate that are aligned to the hard mask (cell or device areas). A second insulating layer is deposited to a thickness sufficient to fill the shallow trenches and is polished back to the hard-mask layer to form shallow trench isolation and to expose the hard-mask surface. The hard-mask layer is selectively removed, such as by wet etching in a hot phosphoric acid solution. This results in recesses in the field oxide isolation that are self-aligned over the device areas and also exposes the first insulating (stress-release) layer in the recesses. Next, openings are etched in the first insulating layer over the device areas to expose the substrate. For example, the bit line contact mask can be used to etch the openings, thereby saving additional mask cost. Next, an epitaxial layer is selectively grown from the silicon substrate exposed in the openings and extends laterally over the first insulating layer in the recesses. By the method of a first embodiment, a portion of the epitaxial layer is doped N⁺ over the first insulating layer to form capacitor bottom electrodes in regions where flat capacitors are to be formed for the DRAM cells. A thin gate oxide is formed on the epitaxial layer, for example by thermal oxidation. A polysilicon layer is deposited on the substrate and is doped N⁺ by ion implantation. The polysilicon layer is then patterned to form FET gate electrodes over the openings in the first insulating layer and also to form capacitor top electrodes for the capacitors over the capacitor bottom electrodes. The FET thin gate oxide also serves as an interelectrode dielectric layer for the flat capacitor. In addition, the polysilicon layer can be concurrently patterned to form polysilicon resistors on the shallow trench isolation. Lightly doped source/drain areas are formed in the epitaxial layer adjacent to the gate electrodes, and insulating sidewall spacers are then formed on the gate electrodes. The DRAM FETs are now completed by forming first and second source/drain contact areas, one on each side of the FET gate electrode adjacent to the sidewall spacers, by ion implantation. The dopant regions in the first source/drain contact areas are contiguous with the doped capacitor bottom electrodes. Bit line electrical contacts are formed to the second source/drain areas to complete the DRAM cells.

In the second embodiment the process is identical to the first embodiment up to and including the deposition of the polysilicon layer to form the gate electrodes. The implant to form the bottom electrodes of the flat capacitors in the first embodiment is optional in the second embodiment, and can be eliminated to reduce process cost. The polysilicon layer is then patterned to form only the FET gate electrodes over the openings in the first insulating layer. The N^- lightly doped source/drain areas in the epitaxial layer are implanted adjacent to the gate electrodes. Insulating sidewall spacers are formed on the gate electrodes, and N⁺ doped first and second source/drain contact areas are formed in the epitaxial layer adjacent to the sidewall spacers by ion implantation to complete the FETs. Continuing with the second embodiment, the stacked capacitors are formed next. A first interpolysilicon oxide (IPO1) layer is deposited, and first contact openings are etched in the IPO1 to the first source/drain contact areas. Capacitor node contacts are formed in the first contact openings, for example by depositing an N⁺ doped polysilicon layer and polishing back. The stacked capacitors are then formed over the node contacts by various means, as commonly practiced in the industry. A second interpolysilicon oxide (IPO2) layer is deposited, and second contact openings for bit lines are etched to the second source/drain contact areas. Conducting plugs are formed in the second openings, and a conducting layer is deposited and patterned to form the bit lines to complete the array of DRAM cells having stacked capacitors. In both embodiments the first insulating layer, utilized as a stress-release layer for the hard-mask layer, is also used under the epitaxial layer. The dual use of the first insulating layer results in reduced process cost while reducing the capacitor leakage current to the substrate.

BRIEF DESCRIPTION OF THE DRAWINGS

The objects and other advantages of the invention will become more apparent in the preferred embodiments when read in conjunction with the following drawings.

FIGS. 1 through 9 show schematically cross-sectional views of one of the memory cells for the sequence of process steps for making the DRAM cell by the first embodiment using the selective epitaxial grown layer on an insulator.

FIG. 10 shows a schematic cross-sectional view of one of the memory cells for the sequence of process steps for making the DRAM cell by the second embodiment using the selective epitaxial grown layer on an insulator.

DESCRIPTION OF THE PREFERRED EMBODIMENTS

The method for making the DRAM cells by a first embodiment using the selective epitaxial silicon layer over an insulating layer is now described in detail. The method and structure are applicable to both simple flat band or stacked capacitor DRAM devices. This novel structure can also be used for transistors, in general, to reduce leakage current.

Referring to FIG. 1, the method by a first embodiment begins by providing a P doped single-crystal silicon substrate 10 having a <100> crystallographic orientation. A first insulating layer 12 comprised of silicon oxide (SiO₂) is formed either by thermal oxidation or by low-pressure chemical vapor deposition-(LPCVD). Layer 12 serves as a stress-release layer for a Si₃N₄ hard mask and is also essential to the current invention as will become obvious at a later step. The SiO₂ first insulating layer 12 is formed to a preferred thickness of between about 100 and 200 Angstroms. A hard-mask layer 14, composed of Si₃N₄, is deposited on the first insulating layer 12. The Si₃N₄ layer 14 is deposited by LPCVD to a preferred thickness of between about 1600 and 2000 Angstroms. Photolithographic techniques and anisotropic plasma etching are used to pattern the hard mask to leave portions over the desired device areas 1. The Si₃N₄ is patterned using reactive ion etching (RIE) and an etchant gas mixture such as CF₄, O₂, and CHF₃, or a mixture such as HBr, SF₆, and O₂. With the photoresist mask (not shown) still in place shallow trenches 2 are plasma etched in the substrate 10. The shallow trenches are plasma etched using RIE and an etchant gas mixture such as Cl₂, HBr, and O₂ to a preferred depth of between about 2000 and 3300 Angstroms. After removing the photoresist mask, for example by plasma ashing in oxygen, a second insulating layer 16 is deposited to a thickness sufficient to fill the shallow trenches 2, and more particularly to a thickness that is at least greater than 6300 Angstroms, as shown in FIG. 1. The second insulating layer 16 is preferably composed of SiO₂, deposited by LPCVD or by high-density plasma deposition, using a reactant gas such as tetraethosiloxane (TEOS)

Referring to FIG. 2, the second insulating layer 16 is chemically-mechanically polished (CMP) back to the hard-mask layer 14 to form the shallow trench isolation 16 and to expose the surface of the hard mask over the device areas 1.

Referring to FIG. 3, the Si₃N₄ hard-mask layer 14 is selectively removed, for example by using a wet etching in a hot phosphoric acid (H₃PO₄) solution at a temperature of about 120 to 200°C C. This results in recesses 3 in the field oxide isolation 16 that have a depth that is equal to the thickness of the Si₃N₄ layer 14. The recesses 3 are self-aligned over the device areas 1 and the first insulating layer 12 is exposed in the recesses.

Referring to FIG. 4, openings 4 are etched in the first insulating layer 12 over the device areas 1 to expose the substrate 10. The openings are etched using a patterned photoresist layer and high-density plasma (HDP) etching using an etchant gas such as CHF₃, C₄F₈, or CH₂F₂ that selectively etches the SiO₂ layer 12 to the substrate. The openings 4 have a diameter or width y that is preferably between about 0.1 and 0.5 micrometers (um). The single-crystal silicon substrate surface in the openings serves as the single-crystal-seed surface having a <100> crystallographic orientation for the epitaxial layer that is grown in the next step. The distance x from the center of the opening 4 to the edge of the shallow trench 16 has a minimum width of about 0.2 um and is sufficiently wide to accommodate the flat capacitor for the DRAM device. For example, the bit line contact mask for the DRAM process can be used to etch the openings 4, thereby saving processing cost.

Referring to FIG. 5, an epitaxial layer 18 is selectively grown from the seed surface of the silicon substrate 10 in the openings 4 and extending laterally over the first insulating layer 12 in the recesses 3. The epitaxial layer 18 is grown in an epitaxial reactor (CVD system) at, high temperature. Typically the selective epitaxial layer 18 is grown using a reactant gas such as SiH₄ or Si₂Cl₂ at a temperature of between about 950 and 1100°C C. The epitaxial layer 18 is doped P type using diborane hydride (B₂H₆) and to a preferred concentration of between about 1.0 E 16 and 1.0 E 18 atoms/cm³. Epitaxial layer 18 is grown to a preferred thickness that is less than the depth of the recess 3 in the field oxide 16, and more specifically to a thickness of about 1000 to 5000 Angstroms. The epitaxial layer 18 is selectively grown on the first insulating layer 12 and if necessary can be wet etched back to the desired thickness.

Referring to FIG. 6, by the method of a first embodiment, a portion of the P doped epitaxial layer 18 is doped N⁺ to form capacitor bottom electrodes 18' in regions where the flat capacitors are to be formed for the DRAM cells. The epitaxial layer is doped by using a photoresist ion implant block-out mask and is implanted with arsenic or phosphorus ions to a final dopant concentration of between about 1.0 E 19 and 4.0 E 21 atoms/cm³. As shown in FIG. 6, the capacitor bottom electrodes 18' are formed over the first insulating layer 12 and away from the opening 4 to reduce capacitor leakage current.

Referring to FIG. 7, a thin gate oxide 20 is formed on the epitaxial layer 18. For example, the oxide 20 is formed by a dry thermal oxidation to a thickness of between about 15 and 35 Angstroms. Alternatively, for future technologies more advanced gate dielectric layers, such as Si₃N₄, TaO_x, and the like can be used, and the thickness of which would be technology-dependent.

Referring to FIG. 8, a blanket polysilicon layer 22 is deposited on the substrate. Layer 22 is deposited by LPCVD using a reactant gas such as silane (SiH₄), and to a thickness of between about 1000 and 2000 Angstroms. The polysilicon layer 22 is then doped N⁺ by ion implanting phosphorus (p³¹) to achieve a final dopant concentration of between about 1.0 E 19 and 4.0 E 21 atoms/cm³. By including additional processing steps, the polysilicon layer 22 can include an upper metal silicide layer and an insulating cap layer, which are not depicted in the FIGS. to simplify the drawings.

Referring to FIG. 9, conventional photolithographic techniques and anisotropic plasma etching are used to pattern the polysilicon layer 22 to form FET gate electrodes 22A over the openings 4 and also to form capacitor top electrodes 22B for the capacitors over the capacitor bottom electrodes 18'. The FET thin gate oxide 20 also serves as an interelectrode dielectric layer 20' for the flat capacitor. In addition, the polysilicon layer 22 can be concurrently patterned to form polysilicon resistors 22C on the shallow trench isolation 16.

Continuing with FIG. 9, lightly doped source/drain areas 17(N^-) are formed in the epitaxial layer 18 adjacent to the gate electrodes 22A. Typically the lightly doped source/drain areas are formed by ion implanting arsenic or phosphorus dopants, preferably arsenic, to achieve a dopant concentration of between about 1.0 E 18 and 2.0 E 20 atoms/cm³. Next, insulating sidewall spacers 24 are formed on the gate electrodes 22A by depositing a conformal insulating layer consisting of a thin SiO_x of about 150 Angstroms and a Si₃N₄ layer having a thickness of between about 500 and 1500 Angstroms, and more specifically a thickness of about 1000 Angstroms. The insulating sidewall spacers 24 are formed by anisotropically etching back. For example, the SiO_x can be deposited by LPCVD using TEOS as the reactant gas, and the Si₃N₄ can be deposited by LPCVD using SiCl₂H₂ and ammonia (NH₃) as the reactant gases. The sidewall spacers 24 are formed by anisotropically etching back the SiO_x/Si₃N₄ layer using RIE and an etchant gas mixture such as CHF₃, CF₄, and O₂. The DRAM FETs are now completed by forming first and second source/drain contact areas 19(N⁺), one on each side of the FET gate electrodes 22A adjacent to the sidewall spacers 24. The source/drain contact areas 19 are formed by implanting arsenic ions to achieve a dopant concentration of between about 1.0 E 19 and 4.0 E 21 atoms/cm³. An important feature of this invention is that the dopant regions in the first source/drain contact areas 19(N⁺) are contiguous with the doped capacitor bottom electrodes 18'.

Still referring to FIG. 9, an interpolysilicon oxide (IPO) layer 26 is deposited to insulate the underlying capacitor and FET devices. Layer 26 is a doped SiO₂ (doped, for example, with boron or phosphorus to a concentration of about 2 to 5%), and is deposited by subatmospheric CVD using, for example, TEOS/0₃ as the reactant gas. Layer 26 is planarized by CMP to have a thickness of between about 5000 and 6500 Angstroms over the capacitor top electrodes 22B. The novel DRAM cell is now completed up to the bit line contact openings by etching the bit line openings 6 in the IPO layer 26 to the second source/drain contact areas 19(N⁺). The bit line openings 6 are etched using conventional photolithographic techniques and anisotropic plasma etching in a high-density plasma etcher.

Referring now to FIG. 10, in the second embodiment the process is identical to the first embodiment up to and including the deposition of the polysilicon layer 22 to form the gate electrodes. Similar elements in the drawings are labeled the same for both embodiments. The polysilicon layer 22 is deposited and patterned to form only the FET gate electrodes 22A over the openings in the first insulating layer 12. The top electrodes of the flat capacitor of the first embodiment are not formed in the second embodiment. The implant to form the bottom electrodes of the flat capacitors in the first embodiment is optional in the second embodiment, and can be eliminated to reduce process cost. The N^- lightly doped source/drain areas 17(N^-) in the epitaxial layer 18 are implanted adjacent to the gate electrodes 22A. Insulating sidewall spacers 24 are formed on the gate electrodes, and N⁺ doped first and second source/drain contact areas 19(N⁺) are formed in the epitaxial layer 18 adjacent to the sidewall spacers 24 by ion implantation to complete the FETs. The process details for forming these elements are the same as in the first embodiment.

Continuing with the second embodiment and still referring to FIG. 10, the stacked capacitors are formed next. A first interpolysilicon oxide (IPO1) layer 26 is deposited. Layer 26 is preferably SiO₂, deposited by plasma-enhanced CVD using, for example, TEOS as the reactant gas. Layer 26 is planarized to have a thickness of between about 6500 and 8500 Angstroms over the FET gate electrodes 22A. First contact openings 8 are etched in the IPO1 layer 26 to the first source/drain contact areas 19(N⁺) using anisotropic plasma etching, as described above. Capacitor node contacts 28 are formed in the first contact openings 8. For example, the node contacts 28 can be formed by depositing an N⁺ doped polysilicon layer and polishing back to the surface of the IPO1 layer 26. The stacked capacitors 30 are thee formed over the node contacts 28 by various means, as commonly practiced in the industry. For example, the stacked capacitors can include cylindrical-shaped, crown-shaped, fin-shaped, and the like, but are not explicitly. depicted in FIG. 10 to simplify the drawing. A second interpolysilicon oxide (IPO2) layer 32 is deposited over the stacked capacitors 30. Layer 32 is deposited and planarized as for IPO1 layer 26, and has a thickness of between about 4000 and 6000 Angstroms over the stacked capacitors. Second contact openings 9 are etched in the IPO layers 32 and 26 to the second source/drain contact areas 19(N⁺) for bit lines. Conducting plugs 34 are formed in the second openings 9. For example, the plugs 34 are preferably formed by depositing a metal, such as aluminum (Al) or tungsten (W), and would include a barrier/adhesion layer such as titanium/titanium nitride (Ti/TiN) to prevent the metal from reacting with the silicon substrate. Next, a conducting layer 36 is deposited and patterned to form the bit lines 36 to complete the array of DRAM cells. The conducting layer 36 is preferably a multilayer of Ti/TiN/AlCu alloy/Ti/TiN. The Ti/TiN layers are deposited to a thickness of about 150 to 300 Angstroms, and AlCu is deposited to a thickness of about 3500 to 4800 Angstroms.

While the invention has been particularly shown and described with reference to the preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made without departing from the spirit and scope of the invention.

Claims

1. Dynamic random access memory (DRAM) cells on and in an epitaxial silicon layer over an insulating layer on a semiconductor substrate comprised of:

a shallow trench isolation around device areas having recesses over and aligned to said device areas;

said insulating layer on said substrate in said recesses, each of said recesses having an opening in said insulating layer to said substrate;

an epitaxial layer in each of said recesses extending from said opening and laterally over said insulating layer;

a gate oxide on said epitaxial layer in each of said recesses;

FET gate electrodes on said gate oxide and over said openings in said insulating layer, and including lightly doped source/drain areas and source/drain contact areas in said epitaxial layer adjacent to said gate electrodes;

capacitor node contacts to said source/drain contact areas in said epitaxial layer over said insulating layer;

bit line contacts in said epitaxial layer over said insulating layer;

capacitors over and contacting said capacitor node contacts, and bit lines over and contacting said bit line contacts.

2. The structure of claim 1, wherein said semiconductor substrate is a P doped single-crystal silicon substrate.

3. The structure of claim 1, wherein said insulating layer is silicon oxide having a thickness of between about 100 and 200 Angstroms.

4. The structure of claim 1, wherein said epitaxial layer is single-crystal silicon and has a thickness of between about 1000 and 5000 Angstroms.

5. The structure of claim 1, wherein said gate oxide is a silicon oxide having a thickness of between about 15 and 35 Angstroms.

6. The structure of claim 1, wherein said gate electrodes are conductively doped polysilicon.